ELECTRICAL CHARACTERIZATION AT ROOM TEMPERATURE

The DC characterization of these devices was performed using HP 4155C parameter analyzer and Keithley 2400 SMU at room temperature. The high temperature DC characterization was performed by heating up the chuck using Temptronic heating module. The Capacitance-Voltage (C-V) measurements were carried out using HP 4284A LCR-meter. The investigation starts by doing C-V characterization of undoped and doped

barrier (i/n)- Al0.85Ga0.15N/ i-Al0.65Ga0.35N HEMTs. Figure 5.2 shows the C-V

characteristics of undoped and doped barrier HEMT. It is found that for UWBG AlGaN based HEMTs the C-V curves followed the similar type of pattern as like conventional

AlGaN/GaN HMETs. Figure 5.2(a) shows that undoped barrier i-Al0.85Ga0.15N/i-

~ - 8V. For the case of doped barrier n- Al0.85Ga0.15N/i-Al0.65Ga0.35N HEMTs shown in

figure 5.2 (b) the zero volt capacitance is ~ 0.30µF/cm2 and it starts depleting at ~-12V

(a) (b)

Figure 5.2 C-V characteristics of undoped (a) and doped (b) barrier HEMTs tested on C-

V pad with area of 2×10-4 cm-2 at room temperature.

(a) (b)

Figure 5.3 (a) Simulated band diagram along the growth direction and (b) 2DEG dependence of barrier thickness for undoped structure.

In both cases the hysteresis is very small even less than 10 mV which is an indication of high quality Schottky contact. From the zero volt capacitance, it seems that doping causes only less than 10% of variation in 2DEG values between the two structures which is

consistent with the results of Rsheet values for doped and undoped barrier HEMTs

explained in chapter 2. 2DEG can be calculated from the C-V characteristics obtained from HP 4284A LCR meter using the following expression,

𝑛_𝑠(𝑉_𝐺) = _𝑞1∫ 𝐶(𝑉_𝑉𝑉_𝑇𝐺 𝐺)𝑑𝑉𝐺 (5.1)

Here the symbols have their usual meanings. The estimated 2DEG density for

undoped barrier HEMT using the above expression is ~1.1×1013cm-2. The simulated band

diagram with electron distribution and 2DEG dependence of barrier thickness are shown in figure 5.3 (a) and 5.3 (b) respectively. Electron distribution obtained from 1-D simulation performed by the Poisson solver [112] is in good agreement with the result obtained from Hg-probe C-V measurement. 2DEG density obtained from simulation is

1.15×1013_cm-2 _{which is in excellent agreement with the estimated 2DEG density from the}

experimentally obtained C-V characteristics. For the doped barrier HEMT the 2DEG density can be lies within 10% range of what is obtained for undoped barrier HEMTs.

The hysteresis of the C-V curves is very negligible (< 10mV) and confirms good quality of Schottky contacts which is an essential requirement for getting low gate

leakage current (IGS). Moreover, high Al content ensures low electron affinity which can

form large Schottky barrier height with Ni (Φm ~ 5eV). Figure 5.4 shows the gate leakage

current measured at room temperature for doped and undoped barrier HEMT and

Figure 5.4 Gate leakage current characteristics of i-Al0.85Ga0.15N/Al0.65Ga0.35N and n-

Al0.85Ga0.15N/Al0.65Ga0.35N HEMTs. Al0.25Ga0.75N/GaN MOSHEMTs data is taken from

ref [104].

From the figure, It is clear that undoped barrier HEMT shows very low gate

leakage current as compare to doped barrier and Al0.25Ga0.75N/GaN MOSHEMTs.

Undoped barrier HEMT exhibits almost 102 times lower gate leakage than doped barrier

and Al0.25Ga0.75N/GaN MOSHEMTs. This is due to large Schottky barrier height which is

around 2.67eV for x > 0.8 in AlGaN [105]. For the doped barrier HEMT, Fermi level pinned at the surface close to conduction band which causes higher leakage current in doped barrier structure. Figure 5.5 shows output characteristics of the undoped and doped barrier HEMTs. From figure 5.5 (a) it is clear that undoped barrier HEMT shows

very nice drain current saturation with peak drain current (IDS) density ~ 40 mA/mm.

Whereas, figure 5.5 (b) shows that the doped barrier HEMT exhibits peak current density of ~250 mA/mm which is the highest reported current in UWBG AlGaN HEMT till to date [42]. The doped barrier HEMT exhibits approximately ×6 times higher drain current than undoped barrier HEMT. The reason may be the surface to channel voltage which changes significantly in doped barrier HEMT. The surface-channel voltage, in turn,

comprises of two main factors: the voltage drop across the cathode contact [42,113] and surface potential drop due to electron injection from the cathode contact [42, 114]. It is believed that much higher current in the doped structure is achieved due to less negative surface charge (compensated by donor dopants) and higher pinch-off voltage (due to

higher total channel charge). In both cases the peak current densities are taken at VGS =

+4V. Large Schottky barrier height allows higher gate driving voltage even without any gate insulator while comparing conventional GaN channel based HEMT and

MOSHEMTs [104]. Beyond VGS = +4V the gate starts leaking that restricts further

increment of gate voltage. In both cases the drain current drooping is negligible. Figure

5.6 shows transfer characteristics with transconductance (GM) for both undoped and

doped barrier HEMTs as a function of gate-source voltage.

(a) (b)

Figure 5.5 (a) IDS-VDS characteristics of undoped barrier HEMT (i-

Al0.85Ga0.15N/Al0.65Ga0.35N) and doped barrier HEMT (n-Al0.85Ga0.15N/Al0.65Ga0.35N) for -

(a) (b)

Figure 5.6 (a) IDS-VGS and transconductance characteristics of undoped barrier HEMT (i-

Al0.85Ga0.15N/Al0.65Ga0.35N) and doped barrier HEMT (n-Al0.85Ga0.35N/Al0.65Ga0.35N) for

VDS = 20V.

(a) (b)

Figure 5.7 (a) Off-state drain current as a function of VDS for VGS= -20V at room

temperature in three terminal measurements. (b) Breakdown voltage data comparison with previous data of University of South Carolina (USC) [31] and recent data by Sandia National Laboratory [40].

In figure 5.6 IDS and GM as a function of VGS for (a) undoped barrier and (b)

doped barrier HEMTs presented are at room temperature. From figure 5.6 (a) extracted

threshold voltage (VT) for undoped barrier HEMT is ~ -7.5 V whereas from figure 5.6 (b)

higher GM than undoped barrier HEMT due to the modification of surface potential by

dopant charge. From the transfer curves shown in figure 5.6 it is estimated that the on to off current ratio for undoped barrier HEMT is higher than doped barrier HEMT by more

than 104 times. It is a clear indication of achieving hard pinch off behavior from undoped

barrier HEMTs. Therefore, the off-state breakdown voltage measurement was performed on these undoped barrier HEMTs. Figure 5.7 (a) shows the three terminal off state

breakdown characteristics for undoped barrier HEMT (i-Al0.85Ga0.35N/Al0.65Ga0.35N).

Figure 5.7(a) shows the three-terminal breakdown voltage for undoped barrier devices with the gate-drain distances of 2.2μm and 9μm.The gate voltage was kept fixed at −20V.

The breakdown measurements were made on un-passivated devices which were immersed in Fluorinert to avoid air breakdown. The drain currents increased from a

starting value of 60 pA to approximately 190 nA (∼1μA/mm) at ~800V for the 9μm gate-

drain spacing device. At this voltage, we observed catastrophic failure due to flashover sparking between the metal electrodes. Same was observed for the 2.2 μm gate-drain spacing device at 200 V. The measured breakdown field is about ten times lower than the value expected for AlGaN material used in this work. However, since the failure was not related to the breakdown in the gate-drain spacing, we believe that with a proper electrode design, encapsulation, edge termination and field plating the breakdown voltages for the high-Al AlGaN channel HEMTs can be > 1kV. Figure 5.7(b) compares the measured three terminal off-state breakdown voltages with the recent data by Sandia National Laboratory [40] and GaN HEMT from USC. It is seen that UWBG AlGaN HEMT can exhibit higher breakdown voltages than GaN HEMT due to its higher critical electric field. It is very attractive feature that for the same off-state breakdown voltage

UWBG AlGaN HEMT size can be reduced than GaN based HEMT which is very promising to achieve higher power density. This is an important factor for cost reduction. also, smaller size is beneficial for high speed RF applications.